Sandwich bots extracted about $29.7 million earlier than prices from Solana trades over an 11-month interval, based on blockchain information supplier Bitquery. It discovered 96% of detected assaults on memecoin platform Pump.enjoyable’s launch curve and its PumpSwap change.
The investigation, masking profitable swaps, counted 6,850,144 assaults by 2,545 bot wallets affecting 2,154,994 distinct wallets between October 30, 2025, and October 7, 2026. These addresses don’t symbolize a rely of particular person individuals: different bots’ trades accounted for 39% of the estimated gross take.
A sandwich locations a bot’s buy and sale round one other dealer’s swap. The primary buy raises the pool worth; the intervening purchaser receives fewer tokens, and the bot sells into the upper worth. The price reaches the dealer via execution, even when the swap completes efficiently.
Broad slippage marked larger publicity
Amongst trades despatched by swap router Jupiter via Pump.enjoyable’s launch curve or PumpSwap, trades permitting 20%–50% slippage had been sandwiched 14 instances as typically as these permitting below 3%, Bitquery discovered.
Slippage tolerance units how far execution can depart from a quote. Jupiter’s documentation describes a minimal acceptable output after making use of that tolerance. A worth transfer past the restrict could cause a swap to fail; a wider restrict permits a worse fill. That creates a tradeoff between finishing a unstable commerce and limiting the worth concession accepted alongside the best way.
The noticed assault charges didn’t rise uniformly with each wider setting. Trades under 0.5% tolerance had been hit extra typically than these at 1%–3%, whereas the 50%-or-more band was hit much less typically than the 20%–50% band.
The slippage comparability is observational and covers Jupiter routes via Pump.enjoyable’s launch curve or PumpSwap. Jupiter carried solely a small share of all sufferer trades within the investigation.
The $29.7 million estimate values SOL on the minute of every assault and doesn’t measure web revenue. In Bitquery’s separate 63-attack charge pattern, bots retained 68% of mixture gross proceeds, and 18 assaults misplaced cash after prices. That pattern doesn’t set up web earnings for the complete interval. The report additionally identifies inconsistent Pump.enjoyable sale-fee recording that impacts its gross estimate by about 3%.
Bitquery used repeated buying and selling patterns to determine bots, excluded spherical journeys starting in a token’s first 10 slots, and experiences gaps in change and block protection. Chain information can’t show a bot noticed an order beforehand, and a few qualifying wallets could also be quick merchants.
Slippage limits deal with the worth a dealer will settle for. A separate analysis preprint submitted September 23, 2026 examined protected order stream, that means transactions supposed to be shielded from front-running, throughout six blockchains. Its authors discovered sandwich patterns extending throughout trades and blocks and concluded these protections can supply weaker ensures than customers anticipate.
